Fog computing introduces a distributed processing capability close to end-users. The proximity of computing to end-users leads to lower service time and bandwidth requirements. Energy consumption is a matter of concern in such a system with a large number of computing nodes. Renewable energy sources can be utilized to lessen the burden on the main power grid and reduce the carbon footprint, but due to fluctuations, effective utilization of renewable energy sources needs proper resource management. In this paper, we deal with properly managing the resources in a fog environment where the fog nodes are equipped with on-site renewable energy. This paper aims to design an efficient mechanism to dynamically dispatch requests among computing nodes and scale frequency and modulation level, based on current workload and the availability of renewable energy sources, to minimize the service time while keeping the renewable energy utilization and stability at a satisfactory level. We state the problem as the design of a controller for a system with time-varying nonlinear state equations. Accordingly, we borrow the Lyapunov optimization technique from the control theory to design the request dispatching mechanisms and prove its asymptotic optimality. We perform extensive simulations to evaluate the effectiveness of the proposed method. Simulation results demonstrate that our proposed method outperforms naive time aware baseline scheme up to 26% and 39%, respectively, in terms of service time and renewable energy utilization.

In this paper, we focus on the convergence of autonomous demand side management (A-DSM) algorithms which are widely discussed in the literature of smart grid. We prove that the Nash equilibrium of these algorithms is not unique, however, the Nash equilibria form a convex set, and each consumer's payoff is the same over this set. Moreover, it is proved that the A-DSM program is convergent to the equilibria set if the consumers take turns in the algorithm. When a large number of consumers participate in the A-DSM program, it takes a long time to converge. Therefore, we modify the A-DSM algorithm such that the consumers can make decision in parallel. The proposed algorithm increases the convergence rate by several folds while the scheduled load profile and the system cost remain very close to those of the original A-DSM algorithm.

As the Internet evolves into a global all-service communication infrastructure, a key consideration is providing quality of service guarantees over IP with efficient resource utilization in a scalable, flexible, and automatic way. In this article we present a virtual network (VN) based architecture for scalable IP service deployment and efficient network resource management. Particularly considering a DiffServ/MPLS III transport network supporting multiple VNs, we propose a dynamic approach for efficient bandwidth sharing among VNs. The bandwidth sharing is service-level-agreement-based; the spare capacity in underloaded VNs is adaptively and efficiently utilized, and SLA compliance for all the VNs involved is always guaranteed.

The deployment of fog computing resources in industrial internet of things (IIoT) is essential to support time-sensitive applications. To utilize resources efficiently, a brand-new request dispatcher is required to sit between the IIoT devices and the pool of fog resources. The need for such a dispatcher stems from the challenges specific to these systems. Firstly, fog-enabled IIoT systems are highly dynamic and distributed. Second, fog nodes are typically power and resource limited. Finally, many IIoT applications feature critical time-sensitivity, referred to as timeliness, and cannot tolerate response delay beyond a specific threshold. This paper proposes an efficient dispatching algorithm to minimize energy consumption and deadline misses while keeping the system stability at a satisfactory level. We leverage Lyapunov Optimization technique to tackle the problem and handle the system dynamics. We perform extensive simulations to verify the effectiveness of the proposed method and provide sensitivity, scalability and model parameter analysis. The simulation results prove the superiority of the proposed method over the state-of-the-art method up to 22% and 10% in terms of average deadline misses and energy consumption, respectively. Further, we perform practical experiments to prove the validity of the proposed method in a real testbed.